The Glycosyltransferase Pathway: An Integrated Analysis of the Cell Metabolome

Nucleotide sugar-dependent glycosyltransferases (UGTs) are critical to the homeostasis of endogenous metabolites and the detoxification of xenobiotics. Their impact on the cell metabolome remains unknown. Cellular metabolic changes resulting from human UGT expression were profiled by untargeted metabolomics. The abundant UGT1A1 and UGT2B7 were studied as UGT prototypes along with their alternative (alt.) splicing-derived isoforms displaying structural differences. Nineteen biochemical routes were modified, beyond known UGT substrates. Significant variations in glycolysis and pyrimidine pathways, and precursors of the co-substrate UDP-glucuronic acid were observed. Bioactive lipids such as arachidonic acid and endocannabinoids were highly enriched by up to 13.3-fold (p < 0.01) in cells expressing the canonical enzymes. Alt. UGT2B7 induced drastic and unique metabolic perturbations, including higher glucose (18-fold) levels and tricarboxylic acid cycle (TCA) cycle metabolites and abrogated the effects of the UGT2B7 canonical enzyme when co-expressed. UGT1A1 proteins promoted the accumulation of branched-chain amino acids (BCAA) and TCA metabolites upstream of the mitochondrial oxoglutarate dehydrogenase complex (OGDC). Alt. UGT1A1 exacerbated these changes, likely through its interaction with the OGDC component oxoglutarate dehydrogenase-like (OGDHL). This study expands the breadth of biochemical pathways associated with UGT expression and establishes extensive connectivity between UGT enzymes, alt. proteins and other metabolic processes.


Introduction
Maintenance of cellular homeostasis relies on a variety of biosynthetic and catabolic pathways, and involves plentiful biomolecules. The glycosylation pathway mediated by UDP-glycosyltransferases (UGTs) is one of these key metabolic processes. It regulates the biological activity of diverse biochemical compounds, by inactivating them through conjugation with glucuronic acid (GlcA) or, less frequently, with glucose (Glc), the sugar moieties of the co-substrates UDP-GlcA or UDP-Glc [1,2]. The glycosylation reaction also leads to an increased polarity of the aglycone, promoting the excretion of the conjugated product through bile and urine [1,2]. In humans, 19 UGT1 and UGT2 enzymes are involved in the elimination of drugs from all classes, toxins and other xenobiotics disturbing various cellular processes. UGTs also regulate the bioavailability and activity of diverse endogenous metabolites including the product of heme catabolism bilirubin, sex steroid hormones, signaling lipids and serotonin [1,[3][4][5]. For example, a number of receptors are activated by substrates of UGT enzymes, including sex steroids and prostaglandins [3,6,7]. These metabolites are involved in signal transduction pathways that control key transcription factors acting as major regulators of gene expression programs [8,9]. Hence, many of the  The splicing events in the UGT1 and UGT2B genes leading to the UGT1A1 and UGT2B7 canonical enzymes (full lines) and alt. proteins (dashed lines) are schematized. The UGT1A1_i2 alternative protein (alt. UGT1A1) possess a truncated C-terminal, resulting from a stop codon located in the exon 5b, whereas the UGT2B7_i8 alternative protein (alt. UGT2B7) has an in-frame insertion of 32 amino acids encoded by the exon 2b. Stable expression and glucuronidation activity of UGT1A1 (C) and UGT2B7 (D) enzymes (enz.) or alternative proteins (alt.) in HEK293 cells. Expression was revealed by immunoblotting of microsomal fractions of UGT-expressing cell models and control cells (Ctr). UGT1A1 and UGT2B7 enzymes displayed enzymatic activity toward typical substrates of these isoenzymes. Assays were conducted with microsomal extracts of each cell model incubated with bilirubin, arachidonic acid (AA), estradiol (E2), 7-ethyl-10-hydroxy-camptothecin (SN- 38) or zidovudine (AZT). Activities are reported in pmol/mg prot/h (right axis), except for bilirubin and AA (1 × 10 3 area/mg prot/h; left axis). No activity was detected in Ctr cells (not shown).
For metabolomics analyses, cell pellets (40 × 10 6 cells) were rinsed with 1 mL ice-cold PBS and centrifuged (5 min, 525× g), flash-frozen on dry ice, then stored at −80 • C until shipment to Metabolon (Morrisville, NC, USA). Five biological replicates at different passages were prepared for each cell model. Metabolomics profiling was conducted by Metabolon Inc. based on ultra-high-performance liquid chromatography-mass spectrometry (UPLC-MS/MS) [33]. Proteins were removed from samples by methanol-induced precipitation and centrifugation. Each sample extract was divided into five fractions for analyses using four methods: two for analysis by two separate reverse phase (RP)/UPLC-MS/MS methods with positive ion mode electrospray ionization (ESI), one for analysis by RP/UPLC-MS/MS with negative ion mode ESI, one for analysis by hydrophilic interaction chromatography (HILIC) UPLC-MS/MS with negative ion mode ESI, and one sample was reserved for backup. Several types of controls were analyzed in concert with the experimental samples. Each reconstitution solvent contained a series of internal standards at fixed concentrations to ensure injection and chromatographic consistency. Raw data were extracted, peak-identified and QC processed using Metabolon proprietary hardware and software. Compounds were identified through comparison with library entries of purified standards or recurrent unknown entities. Data were normalized for protein concentration, as well as log-transformed and median-scaled to attain variables homoscedasticity. Since four groups were compared by UGT subtype, we selected the analysis of variance (ANOVA). p-Values reported throughout the paper were corrected with Tukey's post hoc test for multiple comparisons, as well as for multiple testing using the false discovery rate (FDR) method.

Fatty Acid Synthesis Assay
Fatty acid synthase (FASN) activity was based on a published protocol [37]. Briefly, HEK293 cells from two confluent 15 cm Petri dishes were scraped and rinsed with icecold PBS. After centrifugation (525× g, 5 min), cells were resuspended in lysis buffer (50 mM Tris-HCl pH 7.4, 1 mM EDTA, 150 mM NaCl and complete Protease inhibitor [Sigma]). Cell suspensions were homogenized on ice using a microtip sonicator and a dounce homogenizer. Samples were centrifuged (14,000× g, 15 min, 4 • C) and supernatants were quantified using a Bradford assay. Assays were conducted with 100 µg of protein in an assay buffer (200 mM potassium phosphate buffer pH 6.6, 1 mM DTT, 1 mM EDTA, 240 µM NADPH and 30 µM acetyl-CoA). Malonyl-CoA was added to a final concentration of 50 µM. NADPH oxidation rate was monitored during 10 min at λ = 340 nm with a TECAN M1000 Pro (Morrisville, TN, USA).

Gene Expression Analysis
Flash-frozen cell pellets were obtained as above. RNA extraction was carried out using the RNeasy Plus Mini Kit (Qiagen Inc., Toronto, ON, Canada) as per the manufacturer's protocol. Reverse transcriptase and qPCR were conducted as previously reported [12]. Primer sequences are listed in Table S1. Data were analyzed using the ∆∆Ct method [38].

Validation of Catalytic Properties of Ectopically Expressed UGT Enzymes and Alt. Proteins
We studied the metabolic impact of UGT1A1 and UGT2B7 by stably expressing these enzymes in the UGT-negative cell line HEK293 (Figure 1). The influence of alternative UGT proteins was also studied in the same models, expressed alone or in conjunction with their canonical counterpart. These alternative UGTs are representative of truncated (UGT1A1_i2) and extended (UGT2B7_i8) UGT proteins expressed in several human tissues and perturbed in tumors [12,22]. The shorter alt. UGT1A1_i2 lacks the canonical sequence of 99 amino acids comprising the trans-membrane domain and a short cytosolic charged tail (encoded by exon 5a). This sequence is replaced by a truncated C-terminus encoding a unique sequence of 10 amino acids derived from exon 5b ( Figure 1A). Its molecular weight corresponds to approximately 45 kDa, as confirmed by Western blotting. The longer alt. UGT2B7_i8 has a unique 32-residue in-frame internal region derived from the novel exon 2b residing at the interface between the N-terminal substrate-binding domain and the C-terminal co-substrate-binding domain, leading to an apparent molecular mass of 62 kDa ( Figure 1B).
Immunoblotting and functional assays with known endogenous (bilirubin, estradiol, arachidonic acid) and xenobiotic (anti-cancer agent SN-38 and the anti-viral agent zidovudine) substrates demonstrate the expression of catalytically active UGT enzymes ( Figure 1C,D). In these assay conditions and using UDP-GlcA as the co-substrate, enzyme activity was also observed for the UGT2B7_i8 protein whereas glucuronide formation was not detected for UGT1A1_i2 ( Figure 1C,D), consistent with previous reports [12,17].

The Cellular Metabolome Is Broadly Affected by the Expression of Canonical and Alt. UGT Proteins
An unbiased metabolomics analysis of cell lysates allowed the detection of 615 metabolites, (Tables S2 and S3). Nearly half of the measured metabolites were significantly changed with the expression of UGT enzymes after correction for Tukey's post hoc test for multiple comparisons and for FDR. Compared to control cells, the levels of 276 metabolites were significantly altered in cells expressing the UGT1A1 enzyme (176 increased and 100 decreased). The number of perturbed metabolites reached 345 metabolites in cells expressing the UGT2B7 enzyme with 228 increased and 117 decreased (Table 1, Figure 2A). Nearly all measured metabolite classes were perturbed (Table 2; Figure S1). Compared to control cells, levels of glucose, mannose, fructose, along with several glycolytic intermediates, and 2 -deoxycytidine were the most severely changed (by −100 to 18-fold), but dissimilarly across all UGT proteins (detailed below) ( Table 2). encoding a unique sequence of 10 amino acids derived from exon 5b ( Figure 1A). It molecular weight corresponds to approximately 45 kDa, as confirmed by Western blot ting. The longer alt. UGT2B7_i8 has a unique 32-residue in-frame internal region derived from the novel exon 2b residing at the interface between the N-terminal sub strate-binding domain and the C-terminal co-substrate-binding domain, leading to an apparent molecular mass of 62 kDa ( Figure 1B).
Immunoblotting and functional assays with known endogenous (bilirubin, estradi ol, arachidonic acid) and xenobiotic (anti-cancer agent SN-38 and the anti-viral agen zidovudine) substrates demonstrate the expression of catalytically active UGT enzyme ( Figure 1C,D). In these assay conditions and using UDP-GlcA as the co-substrate, enzym activity was also observed for the UGT2B7_i8 protein whereas glucuronide formation was not detected for UGT1A1_i2 ( Figure 1C,D), consistent with previous reports [12,17].

The Cellular Metabolome Is Broadly Affected by the Expression of Canonical and Alt. UGT Proteins
An unbiased metabolomics analysis of cell lysates allowed the detection of 615 me tabolites, (Tables S2 and S3). Nearly half of the measured metabolites were significantly changed with the expression of UGT enzymes after correction for Tukey's post hoc tes for multiple comparisons and for FDR. Compared to control cells, the levels of 276 me tabolites were significantly altered in cells expressing the UGT1A1 enzyme (176 in creased and 100 decreased). The number of perturbed metabolites reached 345 metabo lites in cells expressing the UGT2B7 enzyme with 228 increased and 117 decreased (Tabl 1, Figure 2A). Nearly all measured metabolite classes were perturbed (Table 2; Figure S1) Compared to control cells, levels of glucose, mannose, fructose, along with several gly colytic intermediates, and 2′-deoxycytidine were the most severely changed (by −100 t 18-fold), but dissimilarly across all UGT proteins (detailed below) ( Table 2).  analysis revealed perturbations common to UGT1A1 and UGT2B7. Displayed pathways were enriched in cell models with an enrichment score > 1 and comprised at least 3 metabolites. (C) Venn diagram of common and divergent changes in metabolite levels (p < 0.05) for UGT1A1 and alt. UGT1A1 models (left). The scatter plot shows the Log2 FC of metabolites altered in both cell models (right). (D) Common and divergent changes in metabolite levels (p < 0.05) for UGT2B7 and alt. UGT2B7 models are displayed in the Venn diagram (left). The scatter plot shows the log2 FC of metabolites altered in both conditions. Pathway enrichment analysis for (E) and (F) are displayed in Figure S1. Cells were cultured in standard conditions, as described in the Methods. Metabolites were categorized according to Metabolon proprietary database. The complete list of metabolites and their quantification are provided in Tables S2 and S3.    Figure 2B). Changes common to UGT1A1 and UGT2B7 enzymes included 190 metabolites, and most notably glucose-6phosphate, mannose-6-phosphate and arachidonic acid (AA) ( Table 2). Alt. UGT proteins also largely affected the cellular metabolome. Alt. UGT1A1_i2 expression modified the levels of 207 cellular metabolites (97 increased and 110 decreased; Table 1, Figure 2C), with 127 metabolites (61%) being also altered in UGT1A1 canonical enzyme-expressing cells. Of those, 94 metabolites (48 increased and 46 decreased) were similarly affected in both cell models, representing 45% and 34% of metabolites altered by alt. and enzyme UGT1A1 expression, respectively. The levels of 292 metabolites (184 increased and 108 decreased) were perturbed by alt. UGT2B7_i8, including 198 metabolites (68%) in common with UGT2B7 enzyme-expressing cells ( Figure 2D). Of these, 147 metabolites were similarly modified (101 increased and 46 decreased). The co-expression of enzymes with their alt. variants induced similar numbers of changes relative to each UGT protein expressed alone (Table 1).

Nucleotide Sugar Precursors Are Significantly Altered by UGT Enzymes and Alt. UGT Proteins
Consistent with the utilization of UDP-GlcA as a co-substrate by UGT canonical proteins, cellular levels of metabolites related to its synthesis were modified in enzymeexpressing cells. For example, several glycolytic intermediates were reduced in enzymeexpressing cells, while metabolites of the pentose phosphate pathway, as well as purines and pyrimidines, were higher. However, metabolites of the hexosamine pathways remained unaffected (Figure 3). More precisely, the levels of orotate, orotidine and uridine-5monophosphate (UMP) metabolites from the pyrimidine synthesis pathway were up to 3.6-fold higher (p < 0.01) in UGT enzyme-expressing cells when compared to control cells ( Figure 3). Expression of the UGT2B7 enzyme also induced elevated levels of N-carbamoyl aspartate (9.0-fold; p = 0.004), a metabolite resulting from a committed step of pyrimidine synthesis. In addition, metabolites from pathways competing with UDP-GlcA synthesis were severely depleted in the presence of UGT enzymes. This included glycolytic and pentose phosphate intermediates such as Glc/Fru-1,6-bisphosphate, glycerate-3-phosphate, phosphoenolpyruvate and 6-phosphogluconate, decreased by −2.6 to −12.5-fold.
Alt. proteins displayed divergent metabolic alterations relative to respective enzymeexpressing cells, with UGT2B17 isoforms showing the most striking differences. Indeed, glycolytic and pyrimidine synthesis intermediates were elevated in cells expressing the alt. UGT2B7 protein, including the co-substrate UDP-GlcA (1.6-fold; p = 0.004) and several glycolytic intermediates (Figure 3). In addition, the alt. UGT2B7_i8 induced a prominent accumulation of glucose (18-fold; p < 0.001) and glucose-6-phosphate (2.3-fold; p < 0.001), an effect that was not modified by its co-expression with the canonical enzyme. Strikingly, co-expression of alt. UGT2B7 mostly reversed the metabolic impacts of the UGT2B7 enzyme ( Figure 3).

UGT1A1 and UGT2B7 Enzymes Strongly Affect the Lipidome and Lipid Metabolism-Related Gene Expression
Higher arachidonic acid (AA) and AA-containing acyl glycerols was a feature of UGT enzyme-expressing cells, in association with significant transcriptional changes in genes encoding AA metabolic enzymes. Indeed, cell models expressing UGT1A1 and UGT2B7 enzymes showed an accumulation of the bioactive lipid AA, with levels increased by 5.4 and 13.3-fold (p < 0.001) in UGT1A1 and UGT2B7 cells, respectively, when compared to control cells (Figures 2A and 4A). Numerous poly-unsaturated fatty acids (PUFAs) and endocannabinoids were also significantly enriched in UGT enzyme-expressing cells when compared to control cells, an effect that was abrogated by co-expression with their alt. isoforms (Table S2). Notably, the monoacylglycerol (MAG) 2-arachidonoylglycerol (2-AG) was significantly higher in UGT1A1 and UGT2B7 enzyme-expressing cells by 5.5 and 2.8-fold (p < 0.001), respectively ( Figure 4A). In line, increased mRNA expression of PLA2G4A, encoding an enzyme releasing AA from phospholipids, and lower expression of MAGL, coding for an enzyme catalyzing MAG hydrolysis, were observed in both UGT enzyme-expressing cells ( Figure 4B). In addition to 2-AG, other endocannabinoid molecules also accumulated in enzyme-expressing cells ( Figure 4C). This was also consistent with a significantly decreased expression of FAAH, which encodes the fatty acid amide hydrolase, the main enzyme catabolizing these bioactive lipids ( Figure 4D).   (Table S2) For UGT1A1-expressing cells, we further observed a decreased expression of genes encoding cannabinoid receptors, including the G-protein coupled receptor CNR1 (−2.9-fold; p < 0.001) and the transient receptor potential cation channel TRPV1 (−1.7-fold; p < 0.001) ( Figure 5A,B). Downstream targets of the cannabinoid system were also decreased, such as mRNA expression (−1.4-fold, p < 0.001) and activity (−1.9-fold, p < 0.01) of the fatty acid synthase (FASN) enzyme, determined by an in vitro assay measuring the oxidation rate of NADPH upon addition of malonyl-CoA ( Figure 5C,D). A decreased expression of endocannabinoid-targeted nuclear receptor PPARD was observed with no significant modifications for PPARA and PPARG. We also perceived a modest but constant decrease in the expression of genes encoding PPAR-regulated mitochondrial and peroxisomal enzymes ( Figure 5A,B), suggesting that the activity of these nuclear recep-  (Table S2) For UGT1A1-expressing cells, we further observed a decreased expression of genes encoding cannabinoid receptors, including the G-protein coupled receptor CNR1 (−2.9-fold; p < 0.001) and the transient receptor potential cation channel TRPV1 (−1.7-fold; p < 0.001) ( Figure 5A,B). Downstream targets of the cannabinoid system were also decreased, such as mRNA expression (−1.4-fold, p < 0.001) and activity (−1.9-fold, p < 0.01) of the fatty acid synthase (FASN) enzyme, determined by an in vitro assay measuring the oxidation rate of NADPH upon addition of malonyl-CoA ( Figure 5C,D). A decreased expression of endocannabinoid-targeted nuclear receptor PPARD was observed with no significant modifications for PPARA and PPARG. We also perceived a modest but constant decrease in the expression of genes encoding PPAR-regulated mitochondrial and peroxisomal enzymes ( Figure 5A,B), suggesting that the activity of these nuclear receptors was repressed. This change in gene expression was more pronounced in cells expressing the UGT1A1 enzyme, as it remained near control levels in cells expressing the alt. UGT1A1 ( Figure 5B).

UGT2B7 Isoforms Affect the Methionine-Creatinine Pathway
For UGT2B7-expressing cells, specific metabolic changes were connected to the creatine pathway. Levels of guanidinoacetate, a metabolite resulting from a committed step of creatine synthesis, were depleted by −3.8 and −1.4-fold (p < 0.05) in cells expressing the enzyme and the alt. UGT2B7 protein, respectively ( Figure 6A). Supporting the induction of this pathway in UGT2B7-expressing cell models, levels of the downstream metabolite creatinine were 1.2-fold higher (p < 0.01) in both cell models when compared to control.

UGT2B7 Isoforms Affect the Methionine-Creatinine Pathway
For UGT2B7-expressing cells, specific metabolic changes were connected to the creatine pathway. Levels of guanidinoacetate, a metabolite resulting from a committed step of creatine synthesis, were depleted by −3.8 and −1.4-fold (p < 0.05) in cells expressing the enzyme and the alt. UGT2B7 protein, respectively ( Figure 6A). Supporting the induction of this pathway in UGT2B7-expressing cell models, levels of the downstream metabolite creatinine were 1.2-fold higher (p < 0.01) in both cell models when compared to control. Elevations of creatine and creatine-P were also noted in cells expressing alt. UGT2B7 alone or together with UGT2B7 enzyme. The production of creatine from guanidinoacetate requires the simultaneous transformation of S-adenosylmethionine (SAM) into S-adenosylhomocysteine (SAH) by guanidinoacetate N-methyltransferase (GAMT). Supporting an increased GAMT activity, SAH levels were enriched by up to 2.0-fold (p < 0.05) in UGT2B7-expressing cells. This is supported by an elevated expression of GAMT by up to 2.0-fold (p < 0.05) in these cells ( Figure 6B). In contrast, few of these metabolites were significantly modified in cells expressing UGT1A1 proteins (Table S2). Supporting an increased GAMT activity, SAH levels were enriched by up to 2.0-fold (p < 0.05) in UGT2B7-expressing cells. This is supported by an elevated expression of GAMT by up to 2.0-fold (p < 0.05) in these cells ( Figure 6B). In contrast, few of these metabolites were significantly modified in cells expressing UGT1A1 proteins (Table S2).  Mitochondrial branched-chain keto acids, i.e., 3-methyl-2-oxovalerate, 4-methyl-2oxopentanoate and 3-methyl-2-oxobutyrate, derived from branched-chain amino acids (BCAA), were more abundant in alt. UGT1A1 expressing cells (alone or co-expressed with the enzyme) by up to 5.9-fold (p < 0.001) when compared to control cells. TCA cycle metabolites, namely citrate, isocitrate and oxoglutarate, were also higher in alt. UGT1A1 expressing cells (by 1.6 to 2.5-fold; p < 0.01; Figure 6C). These changes were exacerbated in cells co-expressing canonical and alt. UGT1A1 proteins, whereas they were not observed in cells expressing the UGT1A1 enzyme alone. By contrast, UGT2B7 enzyme expression led to their depletion, suggesting a differential effect of the UGT proteins on mitochondrial metabolism.
Altered BCAA and TCA cycle metabolites are located upstream of the oxoglutarate dehydrogenase complex (OGDC), among which protein partners of the UGT1As were previously identified by untargeted proteomics experiments in human tissues [10]. While no interaction between the UGT1A1 enzyme and the OGDC component oxoglutarate dehydrogenase-like (OGDHL) protein was detected ( Figure 6D), we observed a proteinprotein interaction between the alt. UGT1A1 and OGDHL, demonstrated in cell models by immunoprecipitation ( Figure 6D,E). This supports the possibility that metabolic changes might be caused by a functional interaction between this member of the OGDC complex and the alt. UGT1A1 protein, explaining the prominent changes of BCAA and TCA cycle metabolites-also linked to FA catabolism-observed in alt. UGT1A1 expressing cells (expressed alone or together with the canonical enzyme).

Discussion
Our study reveals that UGT protein expression triggered significant changes in the cellular metabolome, affecting levels of metabolites in each measured macromolecular group, beyond known UGT substrates. Major changes were observed for both UGT canonical enzymes and their alt. proteins and comprised alterations in carbohydrates, nucleotides and bioactive lipids pathways. Notably, the expression of UGT enzymes induced significant modifications in the metabolism of pyrimidines and glycolysis, suggesting a diversion of these intermediates to support the synthesis of the co-substrates UDP-GlcA and UDP-Glc. Previous lines of evidence implied a limited availability of UDP-GlcA when cells were exposed to important amounts of UGT substrates [40][41][42][43]. An increased pyrimidine metabolism may also reflect a metabolic rewiring to repress UGT activity, as nucleotides represent endogenous allosteric inhibitors [44,45]. Certain UGT enzymes, including UGT2B7, are capable of using UDP-Glc as a co-substrate [46][47][48]. Coherent with a potential increased usage of downstream-related metabolites such as UDP-Glc, depletion in early glycolytic intermediates (e.g., mannose-6-phosphate and glucose-6-phosphate) was observed in canonical UGT2B7 enzyme-expressing cells. Our findings further suggest that UGT enzymes may influence other cellular pathways in which UDP-sugars also participate, including the synthesis of the extracellular matrix component hyaluronan, protein glycosylation, and as ligands of the purinergic receptor P2Y14, which is involved in inflammation, asthma, fibrosis and acute kidney diseases [6,7,49,50].
Notable biochemical routes affected by UGT expression were related to lipid pathways, including the bioactive lipid AA and endocannabinoids. The expression of UGT1A1 and UGT2B7 canonical enzymes induced a significant cellular accumulation of AA levels. This was unanticipated given that AA is a UGT substrate for conjugation [5], and that higher UGT expression would be expected to reduce substrate levels. The accumulation of AA within cells may involve a regulatory feedback loop, as suggested by a previous study reporting the repression of UGT1A1 expression by AA supplementation in the hepatic model HepG2, which endogenously expresses this isoenzyme [51]. It may also engage PPAR signaling, as the lower expression of several peroxisomal and mitochondrial PPAR targets involved in mitochondrial fatty acid beta-oxidation was observed in UGT1A1 enzyme-expressing cells. The interplay between UGT enzymes and lipid homeostasis is also supported by the perturbed accumulation of lipid droplets induced by the expression of UGT1A9 and UGT2B7 in HEK293, breast and pancreas cancer cell models [12,52]. Lipid droplets constitute important storage of energy-rich fatty acids and bioactive lipids that also contribute to limiting lipotoxicity [53,54]. Another report further supports the key role of UGT enzymes in maintaining lipid homeostasis, with an effect on the proliferation of cancer cells [52]. An impact of UGT on gene expression and cell metabolism was also observed in a Ugt1 liver-knock-out mouse model. The loss of Ugt1 functionality in the mouse liver resulted in significant alterations in the expression of several genes including those linked to hormones and fatty acids pathways and pyrimidine metabolism [55]. Accordingly, the interconnection between lipid metabolism and the UGT pathway seems more complex than being related to substrates for these enzymes and likely involves signaling events and modifications of gene expression triggered by bioactive lipids, as well as protein-protein interactions [10,39].
Our study demonstrates that this may also be the case for alt. UGT proteins. For instance, we revealed a protein-protein interaction between the alt. UGT1A1 protein and OGDHL, a component of the OGDC enzyme complex and a key control point in the citric acid cycle often bypassed in cancer cells [56][57][58][59][60]. It is plausible that this partnership may explain, at least in part, changes in the levels of TCA cycle intermediates and BCAA metabolites observed in cells overexpressing the alt. UG1A1 protein. Consistent with this notion, our previous work using an alt. UGT1A-depleted cancer cell model showed a shift in energy metabolism, increasing cell dependency on glucose at the expense of oxidative phosphorylation, likely dependent on a functional interaction of UGT1A proteins with the pyruvate kinase M2 (PKM2) enzyme [10]. These protein partners of the alt. UGT1A1 protein are key regulators of energy metabolism, which could be linked to the capacity of UGTs to induce a redirection of carbon skeletons as discussed above. In fact, when compared to UGT1A1 enzyme-expressing cells, those expressing the alt. UGT1A1 displayed similar alterations for several abovementioned metabolic pathways (glycolysis, pyrimidine synthesis and bioactive lipid metabolism), and could be instigated by common protein interactors [10,11,39]. By these interactions, we also exposed that alt. UGT1A proteins, but not UGT1A enzymes, interfere with oligomeric complex formation necessary for scavenging activity of catalase and peroxiredoxin [11]. Expression of the structurally divergent alt. UGT2B7_i8, yet enzymatically active, caused distinct metabolic profiles for several pathways, notably for glycolytic intermediates and nucleotide sugar precursors, which could also result from protein partners interacting with its unique peptide sequence and/or distinct substrate specificity compared to the canonical UGT2B7 enzyme.
Isoform-specific metabolic and phenotypic changes are likely induced by their divergent primary structure, catalytic function, protein partners and/or subcellular localisation [20]. In line, we previously observed that in contrast to the canonical enzyme, the expression of the alt. UGT2B7_i8 increased cellular adhesion while reducing proliferation, supporting a distinct role for alt. proteins on cellular metabolism [12]. This is also supported by the observation that alt. UGT2B7_i8 expression abrogated many metabolic changes induced by the UGT2B7 enzyme when co-expressed in HEK cells, such as the rewiring of glycolytic intermediates. By contrast, the co-expression of alt. truncated UGT1A1_i2 with its enzyme had pathway-specific effects, such as an exacerbation of the impact of UGT1A1 enzyme on the TCA cycle and repression of its effect on PUFA and MAG. Given their ability to form protein complexes with UGT enzymes [61,62], UGT alt. proteins have the potential to contribute significantly to metabolic changes and those associated with cancer considering their frequent co-expression in cancer tissues. In line, differential isoform usage (isoform switching) is a frequent event in cancer cells, including switches favoring alt. UGT1A1 and UGT2B7 expression that were recently reported in esophageal cancer tissues [26,63].
Because most available immortalized human cancer cell lines express multiple UGTs including enzymes and several forms of alt. UGTs, we selected to conduct this study in the human HEK293 embryonic kidney cell model in which the UGT pathway is inactive due to the lack of endogenous UGT expression ( Figure 1) [62,64,65]. Although this represents a limitation of the study, it permitted outlining the specific effects of individual UGT enzymes and alt. proteins on cellular metabolome. Other limitations include the fact that known signalling molecules inactivated by UGTs are found potentially below quantification in cell models or others may not have been part of the metabolite panel detected by this platform. Moreover, relative metabolite quantification hinders our capacity to compare metabolite levels with other studies, an inherent limitation of untargeted metabolomics analyses. However, our profiling highlighted pathways unsuspected to be related to UGTs, including endocannabinoids. It also revealed direct (substrate for conjugation) and indirect (perturbed gene expression) influences of UGTs on bioactive lipids such as AA. This interplay is also supported by the dependency of UGT activity on interactions with membrane phospholipids [66]. Future work will aim at integrating gene expression, enzyme activity and metabolic perturbations to fully appreciate the connectivity between UGT and other metabolic pathways.
Our study unveiled unprecedented and distinctive changes in intracellular metabolites caused by the expression of two major hepatic canonical UGT enzymes and their prototypical truncated and extended alt. proteins with novel in-frame sequences. Data support that the UGT proteins are involved in a regulatory process used by cells to control the activity of their metabolic networks, with broad consequences on cell metabolite levels linking UGTs to novel metabolic pathways and potential biological functions.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/metabo12101006/s1, Figure S1: Pathway enrichment analysis of metabolic changes common to (A) UGT1A1 canonical and alt. proteins and (B) UGT2B7 canonical and alt. proteins. Figure S2: Full original images of Western blots, related to Figures 1 and 6. Table S1: Primers and conditions for qPCR analysis. Table S2: Levels of measured metabolites in UGT-expressing cells relative to control. Table S3: Levels of measured metabolites per sample.